Execute as Python Script

Run marimo notebooks as standard Python scripts from the command line. This is ideal for automation, batch processing, CI/CD pipelines, and workflows that produce side effects like writing to disk or sending notifications.

Basic Execution

Run any marimo notebook as a Python script:

python my_notebook.py

When executed as a script:

All cells run in dependency order
Outputs go to stdout/stderr
UI elements are not interactive
The script exits when execution completes

Unlike marimo run, which starts a web server, running as a script executes the notebook once and exits. Perfect for cron jobs, automated reports, and data pipelines.

Why Run as a Script?

Use script execution when:

Automating workflows: Scheduled data processing, ETL jobs, report generation
CI/CD pipelines: Testing, validation, building artifacts
Batch processing: Process files, train models, generate outputs
Command-line tools: Interactive CLI applications with argparse
System integration: Call from other programs, shell scripts, or schedulers

Command-Line Arguments

Using argparse

The recommended way to handle arguments uses Python’s built-in argparse:

import marimo as mo
import sys
import argparse

# Define arguments
if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Process dataset")
    parser.add_argument("--input", required=True, help="Input file path")
    parser.add_argument("--output", required=True, help="Output file path")
    parser.add_argument("--verbose", action="store_true", help="Verbose output")
    args = parser.parse_args()
else:
    # Default values when running as notebook
    class Args:
        input = "data/sample.csv"
        output = "results/output.csv"
        verbose = False
    args = Args()

Run it:

python process_data.py --input data.csv --output results.csv --verbose

Using simple-parsing

For more complex configurations, use simple-parsing:

import marimo as mo
from dataclasses import dataclass
import sys

try:
    from simple_parsing import ArgumentParser
except ImportError:
    mo.md("Install simple-parsing: pip install simple-parsing")
    raise

@dataclass
class Config:
    """Training configuration"""
    learning_rate: float = 1e-4
    epochs: int = 100
    batch_size: int = 32
    model_path: str = "model.pkl"

if __name__ == "__main__":
    parser = ArgumentParser()
    parser.add_arguments(Config, dest="config")
    args = parser.parse_args()
    config = args.config
else:
    config = Config()  # Use defaults in notebook

Run it:

python train.py --learning_rate 0.001 --epochs 50 --batch_size 64

Using mo.cli_args()

marimo provides a lightweight argument parser:

import marimo as mo

args = mo.cli_args()
if args:
    dataset = args.get("dataset", "default")
    year = int(args.get("year", 2024))
    debug = args.get("debug", False)
else:
    # Defaults for notebook mode
    dataset = "default"
    year = 2024
    debug = False

Run it:

python script.py -- --dataset sales --year 2024 --debug

mo.cli_args() does basic type inference but doesn’t provide argument validation or help text. For production scripts, use argparse or simple-parsing.

Parameterization Patterns

Environment Variables

Use environment variables for configuration:

import os
import marimo as mo

# Read from environment with defaults
DATABASE_URL = os.getenv("DATABASE_URL", "sqlite:///local.db")
API_KEY = os.getenv("API_KEY", "")
DEBUG = os.getenv("DEBUG", "false").lower() == "true"

if not API_KEY:
    mo.md("⚠️ **Warning**: API_KEY not set").callout(kind="warn")

Run it:

DATABASE_URL=postgresql://prod API_KEY=secret python script.py

Configuration Files

Load parameters from JSON, YAML, or TOML:

import marimo as mo
import json
import sys
from pathlib import Path

if __name__ == "__main__":
    config_path = sys.argv[1] if len(sys.argv) > 1 else "config.json"
else:
    config_path = "config.json"

with open(config_path) as f:
    config = json.load(f)

mo.md(f"Loaded config from {config_path}")

Run it:

python script.py production.json

Conditional Logic

Detect execution mode and adapt behavior:

import marimo as mo
import sys

# Detect if running as script
is_script_mode = __name__ == "__main__"

if is_script_mode:
    # Script mode: minimal output
    def log(msg):
        print(msg)
else:
    # Notebook mode: rich output
    def log(msg):
        return mo.md(msg)

# Use throughout notebook
log("Processing started...")

Output and Side Effects

Writing Files

import marimo as mo
import pandas as pd
from pathlib import Path

# Process data
df = pd.read_csv("input.csv")
results = process(df)

# Write outputs
output_dir = Path("results")
output_dir.mkdir(exist_ok=True)

results.to_csv(output_dir / "processed.csv", index=False)
results.to_parquet(output_dir / "processed.parquet")

mo.md(f"✅ Saved {len(results)} rows to {output_dir}")

Console Output

Print statements appear in the terminal:

import marimo as mo

print("Starting processing...")
for i in range(5):
    print(f"Processing item {i+1}/5")
    # Do work
    
print("✅ Complete!")
mo.md("Processing finished")

Output:

Starting processing...
Processing item 1/5
Processing item 2/5
...
✅ Complete!

Exit Codes

Return meaningful exit codes for automation:

import marimo as mo
import sys

try:
    # Process data
    results = dangerous_operation()
    if not validate(results):
        print("Validation failed!", file=sys.stderr)
        sys.exit(1)
    print(f"Success! Processed {len(results)} items")
    sys.exit(0)
except Exception as e:
    print(f"Error: {e}", file=sys.stderr)
    sys.exit(1)

Check exit codes in bash:

python script.py
if [ $? -eq 0 ]; then
    echo "Success"
else
    echo "Failed"
    exit 1
fi

Integration with Workflows

Cron Jobs

Schedule regular execution:

# Run daily at 2 AM
0 2 * * * cd /path/to/project && python daily_report.py >> logs/$(date +\%Y-\%m-\%d).log 2>&1

GitHub Actions

Integrate with CI/CD:

name: Run Analysis
on:
  schedule:
    - cron: '0 0 * * *'  # Daily at midnight
  workflow_dispatch:

jobs:
  analyze:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - uses: actions/setup-python@v4
        with:
          python-version: '3.11'
      - run: pip install marimo pandas
      - run: python analysis.py --output results.csv
      - uses: actions/upload-artifact@v3
        with:
          name: results
          path: results.csv

Shell Scripts

Orchestrate multiple notebooks:

#!/bin/bash
set -e  # Exit on error

echo "Running ETL pipeline..."

python 01_extract.py --source production
python 02_transform.py --input data/raw --output data/clean
python 03_load.py --target warehouse
python 04_report.py --email team@company.com

echo "✅ Pipeline complete"

Python Subprocess

Call from other Python code:

import subprocess
import sys

result = subprocess.run(
    [sys.executable, "notebook.py", "--", "--param", "value"],
    capture_output=True,
    text=True,
    check=False
)

if result.returncode == 0:
    print("Success:", result.stdout)
else:
    print("Failed:", result.stderr)
    sys.exit(1)

Validation Before Execution

Check notebooks for issues before running:

# Lint notebook
marimo check script.py

# Run only if check passes
marimo check script.py && python script.py

The marimo check command validates:

Multiple definition errors
Delete-nonlocal errors
Cycles in the dependency graph
Other common issues

See the Lint Rules guide for details.

Export with Execution

Combine execution with export to HTML:

# Execute and save HTML output
marimo export html notebook.py -o output.html

# With arguments
marimo export html report.py -o report.html -- --year 2024 --quarter Q1

This runs the notebook and captures all outputs in the HTML file.

Performance Considerations

Optimize script execution:

Cache expensive computations: Use @functools.cache or persist results to disk
Process in chunks: For large datasets, use batch processing
Parallelize: Use multiprocessing or joblib for CPU-bound tasks
Profile first: Use python -m cProfile script.py to identify bottlenecks
Minimize dependencies: Import only what’s needed in each cell

Debugging Scripts

Use Python’s debugger:

# Run with debugger
python -m pdb script.py

# Or add breakpoint in code
import marimo as mo

# Your code
breakpoint()  # Execution pauses here
# More code

Examples

Daily Report Generator

import marimo as mo
import pandas as pd
from datetime import datetime
import argparse

if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("--email", required=True)
    parser.add_argument("--date", default=datetime.now().strftime("%Y-%m-%d"))
    args = parser.parse_args()
else:
    class Args:
        email = "team@example.com"
        date = datetime.now().strftime("%Y-%m-%d")
    args = Args()

# Load and process data
df = pd.read_sql(f"SELECT * FROM sales WHERE date = '{args.date}'", conn)
summary = df.groupby('region')['revenue'].sum()

# Generate and send report
report = mo.md(f"""
# Daily Sales Report - {args.date}
{mo.ui.table(summary)}
""")

if __name__ == "__main__":
    send_email(args.email, report.text)
    print(f"✅ Report sent to {args.email}")

Run daily:

python daily_report.py --email team@company.com

Model Training Pipeline

# Train model with custom parameters
python train_model.py -- --lr 0.001 --epochs 100 --gpu

Data Validation

# Validate data and exit with status code
python validate_data.py --input data.csv || exit 1

Best Practices

For production scripts:

Use argparse for clear argument definitions and help text
Validate inputs before processing
Handle errors gracefully with try/except
Log important events to files or monitoring systems
Return meaningful exit codes (0 for success, non-zero for errors)
Make scripts idempotent so they can safely re-run
Test in notebook mode first before deploying as script

Next Steps

Deploy as App

Run notebooks as interactive web applications

CLI Arguments

Advanced command-line argument handling

Export Formats

Export notebooks to HTML, PDF, and more

CI/CD Integration

Deploy scripts in automated workflows

​Execute as Python Script

​Basic Execution

​Why Run as a Script?

​Command-Line Arguments

​Using argparse

​Using simple-parsing

​Using mo.cli_args()

​Parameterization Patterns

​Environment Variables

​Configuration Files

​Conditional Logic

​Output and Side Effects

​Writing Files

​Console Output

​Exit Codes

​Integration with Workflows

​Cron Jobs

​GitHub Actions

​Shell Scripts

​Python Subprocess

​Validation Before Execution

​Export with Execution

​Performance Considerations

​Debugging Scripts

​Examples

​Daily Report Generator

​Model Training Pipeline

​Data Validation

​Best Practices

​Next Steps

Deploy as App

CLI Arguments

Export Formats

CI/CD Integration

Execute as Python Script

Basic Execution

Why Run as a Script?

Command-Line Arguments

Using argparse

Using simple-parsing

Using mo.cli_args()

Parameterization Patterns

Environment Variables

Configuration Files

Conditional Logic

Output and Side Effects

Writing Files

Console Output

Exit Codes

Integration with Workflows

Cron Jobs

GitHub Actions

Shell Scripts

Python Subprocess

Validation Before Execution

Export with Execution

Performance Considerations

Debugging Scripts

Examples

Daily Report Generator

Model Training Pipeline

Data Validation

Best Practices

Next Steps